We present Schedgehammer, a framework for general-purpose auto-scheduling that optimizes program execution across diverse compiler infrastructures. Unlike existing auto-schedulers that are tightly coupled to specific intermediate representations or rely on template-based search, Schedgehammer introduces a generic, reusable representation of optimization schedules that integrates with multiple user-schedulable languages.

Schedules are modeled as graph-structured objects that capture dependencies between transformations and parameters, enabling systematic mutation, validation, and composition. The framework unifies numerical and structural optimization under a shared interface and supports both random and evolutionary search.

We demonstrate Schedgehammer’s flexibility on TVM and TACO, showing that it effectively optimizes dense and sparse computations while remaining compiler-agnostic. Across benchmarks, it achieves performance comparable to specialized auto-schedulers such as Ansor, highlighting that a unified, extensible abstraction can generalize scheduling beyond individual compiler ecosystems.